Re: oafd lifecycle

[Havoc Pennington <hp redhat com>]

Elliot Lee <sopwith redhat com> writes: 
> Also, a timeout does not distinguish between oafd-as-objectdirectory
> (which needs to timeout), oafd-as-activationcontext (which needs to die
> with the session), and oafd-as-both (the lifetime of which is uncertain).
> 

Yikes. So, I have zero understanding of this - 
 - what is an object directory
 - what is an activation context
 - how do we reliably die with the session

Also, it turns out that pstree shows OAF as child of init, so I'm not
sure what's going on there. (I guess it's that we spawn it via the
double fork.) So now I don't know why oafd is currently dying with the
session...

Ugh, will try to figure it out. I'm also hacking gconfd so that if it
detects another gconfd running, it will pass that old one to oafd and
exit - sort of a "here, you lost this, dumbass" hack. But obviously
this hack shouldn't be propagated to every corba server in GNOME.

Havoc